Function main

examples/dotTracing/dot_tracing.py:192–298

()

def main():
    # Build network.
    network = Network(dt=dt)

    # Input Layer
    inpt = Input(n=dim * dim, shape=[1, 1, 1, dim, dim], traces=True)

    # Hidden Layer
    middle = LIFNodes(n=neurons, traces=True)

    # Ouput Layer
    out = LIFNodes(n=moveChoices, refrac=0, traces=True)

    # Connections from input layer to hidden layer
    inpt_middle = Connection(source=inpt, target=middle, wmin=0, wmax=1)

    # Connections from hidden layer to output layer
    middle_out = Connection(
        source=middle,
        target=out,
        wmin=0,  # minimum weight value
        wmax=1,  # maximum weight value
        update_rule=MSTDPET,  # learning rule
        nu=1e-1,  # learning rate
        norm=0.5 * middle.n,  # normalization
    )

    # Recurrent connection, retaining data within the hidden layer
    recurrent = Connection(
        source=middle,
        target=middle,
        wmin=0,  # minimum weight value
        wmax=1,  # maximum weight value
        update_rule=PostPre,  # learning rule
        nu=1e-1,  # learning rate
        norm=5e-3 * middle.n,  # normalization
    )

    # Add all layers and connections to the network.
    network.add_layer(inpt, name=LAYER1)
    network.add_layer(middle, name=LAYER2)
    network.add_layer(out, name=LAYER3)
    network.add_connection(inpt_middle, source=LAYER1, target=LAYER2)
    network.add_connection(middle_out, source=LAYER2, target=LAYER3)
    network.add_connection(recurrent, source=LAYER2, target=LAYER2)
    network.to(DEVICE)

    # Add monitors
    # network.add_monitor(Monitor(network.layers["Hidden"], ["s"], time=granularity), "Hidden")
    # network.add_monitor(Monitor(network.layers["Output"], ["s"], time=granularity), "Output")
    spikes = {}
    for layer in set(network.layers):
        spikes[layer] = Monitor(
            network.layers[layer],
            state_vars=["s"],
            time=int(granularity / dt),
            device=DEVICE,
        )